How to diagnose a healthcare operational problem before choosing a solution

Most healthcare improvement work does not fail at implementation. It fails earlier, quietly, at the point where a team decides what the problem is. A department under pressure rarely has the luxury of a slow start, so the first plausible explanation becomes the working diagnosis and the project is built around it. Six months later the dashboard looks better, the pressure has not moved, and everyone is slightly less willing to try again.

Diagnosis is the stage that decides whether everything that follows is worth doing. This article sets out how we approach it, why the obvious explanation is usually incomplete, and what a good diagnostic phase produces.

Symptoms are not the problem

Rising locum spend, long waits in SDEC, cancelled clinics, high sickness rates: these are symptoms. They are what a service experiences, and they are usually what triggers the request for help. But a symptom tells you very little about cause. High locum spend can come from genuine vacancies, from a rota template that does not match demand, from leave that is approved without visibility of cover, from recruitment barriers, or from all of these reinforcing each other.

If you treat the symptom as the problem, you will design an intervention that matches the symptom. Booking locums faster is a reasonable response to "we have gaps". It is the wrong response to "our rota is not designed around the service we actually run".

Three sources, triangulated

No single source of information is sufficient for a diagnosis. We combine three, and we are suspicious of any conclusion that only one of them supports.

1. Data

Activity, staffing, leave, temporary staffing spend, patient flow. The point is not volume but shape: how demand varies by hour, day and season; where staffing does and does not match it; how often the same gaps recur. Averages hide most of what matters. A service staffed correctly "on average" can still be dangerously short every Monday evening.

2. Frontline insight

The people who run the service every day already know a great deal about why it does not work. Doctors, nurses, rota coordinators and service managers can usually describe the workarounds, the informal rules and the points where the official process is abandoned. This knowledge rarely appears in data and is almost never written down. Structured conversations and observation surface it.

3. System analysis

How the parts fit together. Which decisions depend on which information, who owns each step, where handoffs happen and what incentives are at work. Many operational problems are not caused by any individual part failing but by the way the parts interact. A rota, a leave policy and a recruitment process can each be sensible in isolation and still produce chronic gaps together.

Common diagnostic traps

• Solution-first thinking: the project begins with a product, a restructure or a new post already decided, and the diagnosis is shaped to justify it.
• Measuring what is easy: using whatever data is already reported rather than what the question needs.
• Treating variation as noise: smoothing over the daily and weekly patterns that are the problem.
• Mistaking a single cause for the whole story: most persistent operational problems are self-reinforcing cycles with several entry points.
• Skipping the baseline: without a clear measured starting point, nobody can later say whether anything changed.

What a good diagnostic phase produces

The output of diagnosis is not a report for the shelf. It is a small set of things that make the next stages possible:

1. 1A problem statement that frontline staff and senior leaders both recognise as true.
2. 2An explanation of root causes, including how they interact.
3. 3A measured baseline: the handful of indicators that will tell you whether change has happened.
4. 4A view on which causes are addressable now, which need longer-term work and which are outside the service’s control.

In our work at Newham University Hospital, the presenting symptom was rising reliance on resident doctor locums. The diagnosis was that the rota was not structured around the department’s real needs and that the processes around it offered little visibility or control. Those are different problems, and they call for different responses. The redesign that followed started with the rota and the service, not with software.

Diagnosis is the first stage, not a delay

It can feel like a diagnostic phase postpones action. In practice it is the fastest route to action that works, because it removes the months spent implementing the wrong thing. It is the D in our DRIVE approach for a reason: everything downstream, from redesign to embedding, depends on getting it right.